a

r
X
i
v
:
1
2
0
1
.
0
7
7
5
v
4


[
q
u
a
n
t
-
p
h
]


1
4

F
e
b

2
0
1
2
On the Origins of Quantum
Correlations
Joy Christian

Wolfson College, University of Oxford,

Oxford, OX2 6UD, United Kingdom
Abstract: It is well known that quantum corre-
lations are not only more disciplined (and hence
stronger) compared to classical correlations, but
they are more disciplined in a mathematically
very precise sense. This raises an important phys-
ical question: What is responsible for making
quantum correlations so much more disciplined?
Here we explain the observed discipline of quan-
tum correlations by identifying the symmetries
of our physical space with those of a parallelized
7-sphere. We substantiate this identication by
proving that any quantum correlation can be un-
derstood as a classical, local-realistic correlation
among a set of points of a parallelized 7-sphere.

jjc@alum.bu.edu
Chapter 1
On the Local-Realistic Origins of
Quantum Correlations
In a good mystery story the most obvious
clues often lead to the wrong suspects.
Einstein and Infeld
1.1 Introduction
In 1927, when quantum theory was still in its infancy and John Bell
was yet to be born, Albert Einsteinone of the founding fathers of
the theorywas attending the now famous 5
th
Solvay Conference
in Brussels. He was profoundly disturbed by what the new theory
had to say about the nature of physical reality. Among other things,
his concerns stemmed from a deep appreciation of unity in nature.
Beyond the cliche of God does not play dice, he had recognized
that quantum theory entailed a fundamental schism in nature. The
aphorism God does not split reality perhaps better captures the
true essence of his concerns [1]. What he sought was a unied picture
of nature, devoid of any subjective boundary between the classical
and the quantum. What he suspected was a deeper layer of reality,
beyond the polarized picture oered by quantum theory.
By 1935when John Bell was seventhese worries of Einstein
had matured into a powerful logical argument against the new theory.
Published in collaboration with Boris Podolsky and Nathan Rosen,
this argument proves, once and for all, that quantum theory provides
at best an incomplete description of the physical reality [2]. Since
the argument itself is logically impeccable, this conclusion is beyond
1
dispute. Any argument, however, can only be as good as its premises,
and thatas is well knownis where Bell entered the game in 1964
[3]. He attempted to show that not all of the premises of EPR are
mutually compatible. Ironically, however, it is the argument of Bell
that turns out to contain a faulty assumption, not that of EPR.
What is more, this assumption appears in the very rst equation of
Bells famous paper [3], and yet it had escaped notice until recently.
In a series of papers, written between 2007 and 2011, I tried to
bring out Bells error and constructed explicit counterexamples, not
only to his original theorem, but also to several of its variants. This
book is a collection of these papers, each of which can be read more
or less independently, but their contents are interconnected, and
reveal dierent aspects of the fundamental aw in Bells argument.
The collection as a whole, however, is better viewed as addressing a
very important physical question. Regardless of the validity of his
theorem, what Bell discovered in 1964 is physically quite signicant.
He discovered that quantum correlations are far more disciplined
than any classically possible correlation. What is more, quantum
correlations are not only more disciplined, but are more disciplined
in a mathematically very precise sense. This tells us something much
more profound about the structure of the world we live in. And, at
the same time, it raises a very important physical question:
What is it that makes quantum correlations
more disciplined than classical correlations?
My goal in this book is to answer this question in mathematically and
physically as precise a sense as possible. To this end, let me begin
with an extended summary of my argument against Bells theorem.
As noted above, the story began with Einstein, Podolsky, and
Rosen [2]. The logic of their argument can be summarized as follows:
(1) QM = Perfect Correlations
+ (2) Adherence to Local Causality
+ (3) Criterion of Objective Reality
+ (4) Notion of a Complete Theory
=(5) QM is an Incomplete Theory.
Given their premises, the conclusion of EPR follows impeccably.
Among their premises (which are hardly unreasonable), the one that
concerns us the most is their criterion of completeness:
every element of the physical reality must
have a counterpart in the physical theory.
2
S
2
The point Bell missed
one-point compactication
IR
2

=
Figure 1.1: Although lines and planes contain the same number
of points, it is impossible to put the points of a line or a plane in
a one-to-one correspondence with all of the points of a 2-sphere.
Bell attempted to prove that no theory satisfying this criterion can
be locally causal. To this end, he took a complete theory to mean
any theory whose predictions are dictated by functions of the form
A(n, ) : IR
3
S
0
1, +1, (1.1)
where IR
3
is the space of 3-vectors, is a space of complete states,
and S
0
1, +1 is a unit 0-sphere. He then claimed (correctly, as
it turns out) that no pair of functions of this form can reproduce the
correlations for the singlet state predicted by quantum mechanics
1
:
A(a, ) B(b, ) ,= a b. (1.2)
At rst sight, this appears to be a straightforward mathematical
contradiction undermining the force of the EPR argument [3]. And
for this reason functions of this form are routinely assumed in the
Bell literature to provide complete specications of the elements of
physical reality, or complete accounting of all possible measurement
results. As we shall see however, Bells prescription is not only false,
it is breathtakingly nave and unphysical. It stems from an incorrect
underpinning of both the EPR argument and the actual topological
congurations involved in the relevant experiments [4]. In truth, for
any two-level system the EPR criterion of completeness demands
that the correct functions must necessarily be of the form
1 = A(n, ) : IR
3
S
3
IR
4
, (1.3)
1
This is hardly surprising. After all, the product moment correlation coecient
employed by Bell in his paper, by denition, is a measure of linear relationship
between bivariate variables. Thus Bell implicitly assumed a linear relationship
between A and B to prove that the relationship between them must be linear!
3
with the simply-connected codomain S
3
of A(n, ) replacing the
totally disconnected codomain S
0
assumed by Bell. It is important to
note here that this correction does not aect the actual measurement
results. For a specic vector n and an initial state we still have
A(n, ) = +1 or 1 (1.4)
as demanded by Bell, but now the topology of the codomain of the
function A(n, ) has changed from a 0-sphere to a 3-sphere, with the
latter embedded in IR
4
in such a manner that the above constraint is
satised. On the other hand, as is evident from Fig. (1.1) (and will be
further claried in the following pages), without this topological cor-
rection it is impossible to provide a complete account of all possible
measurement results. Thus the selection of the codomain S
3
IR
4
in equation (1.3) is not a matter of choice but necessity. What is
responsible for the EPR correlations is the topology of the set of all
possible measurement results [4]. And for a two-level system this set
happens to be an equatorial 2-sphere within a parallelized 3-sphere.
But once the codomain of the functions A(n, ) is so corrected, the
proof of Bells theorem (as given in Ref. [3]) simply falls apart. In
fact, as we shall repeatedly see in the following pages, the strength
of the correlation for any physical system is entirely determined by
the topology of the codomain of the local functions A(n, ). It has
nothing whatsoever to do with entanglement or nonlocality.
1.2 Local Origins of the EPR-Bohm Correlations
Put dierently, once the measurement results are represented by
functions of the form (1.3), it is quite easy to reproduce the quantum
correlations purely classically, in a manifestly local-realistic manner.
For example, suppose Alice and Bob are equipped with the variables
A(a, ) = a
j

j
a
k

k
() =
_
+1 if = +1
1 if = 1
(1.5)
and
B(b, ) = +b
k

k
b
j

j
() =
_
1 if = +1
+1 if = 1 ,
(1.6)
where the repeated indices are summed over x, y, and z; the xed
bivector basis
x
,
y
,
z
is dened by the algebra

k
=
jk

jkl

l
; (1.7)
4
n

n
Figure 1.2: A unit bivector represents an equatorial point of a unit,
parallelized 3-sphere. It is an abstraction of a directed plane segment,
with only a magnitude and a sense of rotationi.e., clockwise () or
counterclockwise (+). Neither the depicted oval shape of its plane,
nor its axis of rotation n, is an intrinsic part of the bivector n.
andtogether with
j
() =
j
the -dependent bivector basis

x
(),
y
(),
z
() is dened by the algebra

k
=
jk

jkl

l
, (1.8)
where = 1 is a fair coin representing two alternative orientations
of the 3-sphere
2
,
jk
is the Kronecker delta,
jkl
is the Levi-Civita
symbol, and a = a
j
e
j
and b = b
j
e
j
are unit vectors [5]. Evidently,
the variables A(a, ) and B(b, ) belonging to S
3
in addition to
being manifestly realisticare strictly local variables. In fact, they
are not even contextual [6]. Alices measurement resultalthough
it refers to a freely chosen direction adepends only on the initial
state ; and likewise, Bobs measurement resultalthough it refers
to a freely chosen direction bdepends only on the initial state .
In the subsequent chapters we shall mainly use the standard
notations of Cliord algebra Cl
3,0
. The bivector algebras (1.7) and
(1.8) will then be seen as even subalgebras of Cl
3,0
. The latter is a
linear vector space, IR
8
, spanned by the graded orthonormal basis
1, e
x
, e
y
, e
z
, e
x
e
y
, e
y
e
z
, e
z
e
x
, e
x
e
y
e
z
, (1.9)
2
Needless to say, A(a, ) and B(b, ) are two dierent functions of the random
variable . Moreover, they are statistically independent events occurring within
a 3-sphere, with factorized joint probability P(A and B) = P(A) P(B)
1
2
.
Therefore their product AB is guaranteed to be equal to 1 only for the case
a = b. For all other a and b, AB will alternate between the values 1 and + 1.
5
where is the outer product, and the trivector I e
x
e
y
e
z
denes the fundamental volume form of the physical space. In terms
of these notations we can rewrite the bivector a
j

j
() as
a a
j

j
() a
x
e
y
e
z
+ a
y
e
z
e
x
+ a
z
e
x
e
y
,
(1.10)
with = I now representing the hidden variable of the model. The
variables A(a, ) and B(b, ) dened above then take the form
S
3
A(a, ) = (I a) ( + a) =
_
+1 if = +I
1 if = I
(1.11)
and
S
3
B(b, ) = (+Ib) ( +b) =
_
1 if = +I
+1 if = I,
(1.12)
with the trivector being either +I or I with equal probability. In
what follows we shall view the xed bivectors (I a) and (+I b)
as representing the measuring instruments for detecting the random
bivectors ( + a) and ( + b), which represent the spins.
It is crucial to note that the variables A(a, ) and B(b, ) are
generated with dierent bivectorial scales of dispersion (or dierent
standard deviations) for each direction a and b. Consequently, in
statistical terms these variables are raw scores, as opposed to stan-
dard scores [7]. Recall that a standard score indicates how many
standard deviations an observation or datum is above or below the
mean. If x is a raw (or unnormalized) score and x is its mean value,
then the standard (or normalized) score, z(x), is dened by
z(x) =
x x
(x)
, (1.13)
where (x) is the standard deviation of x. A standard score thus
represents the distance between a raw score and the population mean
in the units of standard deviation, and allows us to make comparisons
of raw scores that may have come from very dierent sources. In
other words, the mean value of the standard score itself is always
zero, with standard deviation unity. In terms of these concepts the
bivariate correlation between raw scores x and y is dened as
c(x, y) =
lim
n1
_
1
n
n

i =1
(x
i
x ) (y
i
y )
_
(x) (y)
(1.14)
= lim
n1
_
1
n
n

i =1
z(x
i
) z(y
i
)
_
. (1.15)
6
It is vital to appreciate that covariance by itselfi.e., the numerator
of equation (1.14) by itselfdoes not provide the correct measure of
association between the raw scores, not the least because it depends
on dierent units and scales (or dierent scales of dispersion) that
may have been used (advertently or inadvertently) in the measure-
ments of such scores [7]. Therefore, to arrive at the correct measure
of association between the raw scores one must either use equation
(1.14), with the product of standard deviations in the denominator,
or use covariance of the standardized variables, as in Eq. (1.15).
These basic statistical concepts are crucial for understanding the
EPR correlations. As dened above, the random variables A(a, )
and B(b, ) are products of two factorsone random and another
non-random. Within A(a, ) the factor a
k

k
() is a random
factora function of the hidden variable , whereas a
j

j
is a
non-random factor, independent of the hidden variable . Thus, as
a random variable each number A(a, ) and B(b, ) is generated
with a dierent standard deviationi.e., a dierent size of typical
error. More specically, A(a, ) is generated with the standard
deviation a
j

j
, whereas B(b, ) is generated with a dierent
standard deviation, namely +b
k

k
. These two deviations can be
calculated easily. Since errors in linear relations propagate linearly,
the standard deviation of A(a, ) is equal to a
j

j
times the
standard deviation of a
k

k
() (which we write as (A)), whereas
the standard deviation of B(b, ) is equal to +b
k

k
times the
standard deviation of b
j

j
() (which we write as (B)):
(A ) = a
j

j
(A)
and (B) = +b
k

k
(B). (1.16)
But since all the bivectors we have been considering are normalized
to unity, and since the mean of a
k

k
() vanishes on the account
of being a fair coin, its standard deviation is easy to calculate, and
it turns out to be equal to unity:
(A) =

_
1
n
n

i =1

A(a,
i
) A(a,
i
)

2
=

_
1
n
n

i =1
[[ a
k

k
(
i
) 0 [[
2
= 1, (1.17)
where the last equality follows from the fact that a
k

k
(
i
) are
normalized to unity. Similarly, we nd that (B) is also equal to 1.
As a result, the standard deviation of A(a, ) works out to be equal
7
S
3
4D Ball
t

A
B
Figure 1.3: An initial EPR state originated at time t = 0 evolves
into measurement results A and B at a later time, occurring at two
spacelike separated locations on a parallelized 3-sphere, S
3
, which
can be thought of as a boundary of a 4-dimensional ball of radius t.
In what follows we shall assume that t has been normalized to unity.
to a
j

j
, and the standard deviation of B(b, ) works out to
be equal to +b
k

k
. Putting these two results together, we arrive
at the following standardized scores corresponding to the raw scores:
A(a, ) =
A(a, ) A(a, )
(A )
=
A(a, ) 0
a
j

j

= a
k

k
() (1.18)
and B(b, ) =
B(b, ) B(b, )
(B)
=
B(b, ) 0
+b
k

k

= b
j

j
() , (1.19)
where we have used the identities such as +a
k

k
a
k

k
= +1.
Not surprisingly, just like the raw scores A(a, ) and B(b, ), these
standard scores are also strictly local variables: A(a, ) depends only
on the freely chosen local direction a and the common cause , and
likewise B(b, ) depends only on the freely chosen local direction b
and the common cause . Moreover, despite appearances, A(a, )
and B(b, ) are simply binary numbers, 1, albeit occurring within
8
the compact topology of the 3-sphere rather than the real line:
S
3
S
2
A(a, ) = a
k

k
() = 1 about a IR
3
, (1.20)
S
3
S
2
B(b, ) = b
j

j
() = 1 about b IR
3
. (1.21)
In fact, since the space of all bivectors a
k

k
() is isomorphic
to the equatorial 2-sphere contained within the 3-sphere [4], each
standard score A(a, ) of Alice is uniquely identied with a denite
point of this 2-sphere, and likewise for the standard scores of Bob.
In the following chapters we shall tacitly assume
that this procedure of standardizing from the raw
scores to standard scores has been performed for
all measurement results, taken either as points of a
3-sphere, or more generally as points of a 7-sphere.
Now, since we have assumed that initially there was 50/50 chance
between the right-handed and left-handed orientations of the 3-sphere
(i.e., equal chance between the initial states = +1 and = 1),
the expectation values of the local outcomes vanish trivially. On the
other hand, as discussed above, to determine the correct correlation
between the joint observations of Alice and Bob we must calculate
covariance between the corresponding standard scores A(a, ) and
B(b, ), not the raw scores themselves. The correlation between the
raw scores A(a, ) and B(b, ) then works out to be
c(a, b) = lim
n1
_
1
n
n

i =1
A(a,
i
) B(b,
i
)
_
(1.22)
= lim
n1
_
1
n
n

i =1
_
a
j

j
(
i
)
_ _
b
k

k
(
i
)
_
_
(1.23)
= a
j
b
j
lim
n1
_
1
n
n

i =1
_

jkl
a
j
b
k

l
_
_
(1.24)
= a
j
b
j
+ 0 = a b, (1.25)
where we have used the algebra dened in (1.8). Consequently, when
the raw scores A = 1 and B = 1 are compared in practice by
coincidence counts [8][9], the normalized expectation value of their
product will inevitably yield
c(a, b) =
_
C
++
(a, b) + C

(a, b) C
+
(a, b) C
+
(a, b)
_
_
C
++
(a, b) + C

(a, b) + C
+
(a, b) + C
+
(a, b)
_
= a b, (1.26)
9
1
1
0 180 90
+
c
0

Figure 1.4: Local-realistic correlations can be stronger within S
3
.
where C
+
(a, b) etc. represent the number of joint occurrences of
detections +1 along a and 1 along b etc.
The above equation simply describes covariance of the numbers
A = 1 and B = 1 yielding the impossible strong correlation:
c(a, b) = lim
n1
_
1
n
n

i =1
A(a,
i
) B(b,
i
)
_
= a b. (1.27)
How is this impossible result possible? After all, Bell seems to have
proved mathematically [3] that correlation between such numbers
can never exceed the linear limit. The answer is that, apart from the
statistical dierences we discussed above, there are also topological
dierences between the above expression and what Bell considered
in his theorem. In particular, in the above equation the numbers
A = 1 and B = 1 are points of a parallelized 3-sphere,
1 = A(n, ) : IR
3
S
3
IR
4
, (1.28)
and since a parallelized 3-sphere is a compact, simply-connected
topological space, it should not be surprising that correlation among
its points is stronger-than-linear. By contrast in Bells argument the
measurement functions are implicitly assumed to be of the form
1 = A(n, ) : IR
3
J IR, (1.29)
and since the topology of the real line is far less disciplined than
that of a 3-sphere, it is impossible to generate strong correlation
among its points. In other words, the strength of the correlation is
10
entirely dependent on how disciplined the codomain of the functions
A(n, ) is. Now it turns out that the parallelized spheres S
3
and S
7
have maximally disciplined topology in this respect [4]. And since
the 3-sphere can be parallelized by unit quaternions, the covariance
between its equatorial points represented by pure quaternions (or
unit bivectors) is precisely the EPR correlation
c(a, b) = lim
n1
_
1
n
n

i =1
A(a,
i
) B(b,
i
)
_
= a b, (1.30)
where the bivectors A(a, ) = a
k

k
() and B(b, ) = b
k

k
()
are the standardized variables in the statistical terms discussed above.
Then, for arbitrary four directions a, a

, b, and b

, the corresponding
CHSH string of expectation values immediately gives
[c(a, b) + c(a, b

) + c(a

, b) c(a

, b

)[
2
_
1 (a a

) (b

b) 2

2 , (1.31)
which squarely contradicts Bells theorem and exactly reproduces
the quantum mechanical prediction. So much for Bells theorem.
1.3 Local Origins of ALL Quantum Correlations
What the above example shows is that the correlation among the
raw scores observed by Alice and Bob are entirely determined by the
topology of the codomain of the local functions A(a, ) and B(b, ).
In particular, correlations among the points of a unit parallelized
3-sphere are stronger than those among the points of the real line.
Moreover, as we shall see in the following chapters, once the topology
of the codomain is correctly specied, not only the EPR correlations,
but also the correlations predicted by the rotationally non-invariant
quantum statessuch as the GHZ states and Hardy statecan be
exactly reproduced in a purely local-realistic manner. Thus, contrary
to the widespread belief, the correlations exhibited by such states
are not some irreducible quantum eects, but purely local-realistic,
topological eects. In the cases of EPR and Hardy states the correct
topology of the codomain is a parallelized 3-sphere, and consequently
the correlations exhibited by these states are classical correlations
among the points of a parallelized 3-sphere. In the case of GHZ states
on the other hand the correct topology of the codomain is a paral-
lelized 7-sphere, and consequently the correlations exhibited by these
states are classical correlations among the points of a parallelized
7-sphere. More generally, all quantum mechanical correlations can
11
be understood as purely classical, local-realistic correlations among
the points of a parallelized 7-sphere. Needless to say, this vindicates
Einsteins suspicion that a quantum state merely describes a statisti-
cal ensemble of physical systems, not an individual physical system.
What is more, as we shall see in Chapter 7 in greater detail, there
are profound mathematical and conceptual reasons why the topology
of the 7-sphere plays such a signicant role in the manifestation of
all quantum correlations. As is well known, quantum correlations
are more disciplined (or stronger) than classical correlations in a
mathematically precise sense. It turns out that it is the discipline
of parallelization in the manifold of all possible measurement results
that is responsible for the strength of quantum correlations. In fact,
both the existence as well as the strength of all quantum correlations
are dictated by the parallelizability of the spheres S
0
, S
1
, S
3
, and S
7
,
with the 7-sphere being homeomorphic to the most general possible
division algebra. And it is the property of division that is responsible
for maintaining the strict local causality in the world we live in.
These considerations lead us to the following remarkable theorem:
Theorema Egregium:
Every quantum mechanical correlation can be understood
as a classical, local-realistic correlation among a set of points
of a parallelized 7-sphere, represented by maps of the form
1 = A(n, ) : IR
3
S
7
IR
8
. (1.32)
The corresponding physical picture is the same as in Figure 1.3, but
with S
3
replaced by S
7
, the 4D ball replaced by the 8D ball, and
the number of statistically signicant measurement events, A, B,
C, D, etc., generalized to an arbitrary number. It is important to
note, however, that despite appearances neither S
3
nor S
7
is a round
sphere. The Riemann curvature of both S
3
and S
7
is zero, because
they are both parallelized spheres [4].
S
7
, however, has a much richer topological structure than S
3
.
As noted above, it happens to be homeomorphic to the space of unit
octonions, which are well known to form the most general division
algebra possible. In the language of ber bundles one can thus view a
7-sphere as a 4-sphere worth of 3-spheres. Each of its ber is then a 3-
sphere, and each one of these 3-spheres is a 2-sphere worth of circles.
Thus the four parallelizable spheresS
0
, S
1
, S
3
, and S
7
can all
be viewed as nested within a 7-sphere. The EPR-Bohm correlations
can then be understood as correlations among the equatorial points
of one of the bers of this 7-sphere, as we saw above. Alternatively,
a 7-sphere can be thought of as a 6-sphere worth of circles. Thus the
above theorem can be framed entirely in terms of circles, each one of
12
which described by a classical, octonionic spinor with a well-dened
sense of rotation (i.e., whether it describes a clockwise rotation about
a point within the 7-sphere or a counterclockwise rotation). This
sense of rotation in turn denes a denite handedness (or orientation)
about every point of the 7-sphere. If we designate this handedness
by a random number = 1, then local measurement results for
any physical scenario can be represented by raw scores of the form
S
7
A(a, ) = (J N(a) ) (+ N(a) ) =
_
+1 if = +J
1 if = J,
(1.33)
where a IR
3
and N(a) IR
7
are unit vectors, and = J is the
hidden variable analogous to = I with I = e
x
e
y
e
z
replaced by
J = e
1
e
2
e
4
+e
2
e
3
e
5
+e
3
e
4
e
6
+e
4
e
5
e
7
+e
5
e
6
e
1
+e
6
e
7
e
2
+e
7
e
1
e
3
.
(1.34)
The standard scores corresponding to these raw scores are then given
by N(a), which geometrically represent the equatorial points of a
parallelized 7-sphere, just as a represented the equatorial points
of a parallelized 3-sphere. In Chapters 6 and 7 we shall see explicit
examples of the functions N(a) IR
7
corresponding to measurement
directions a IR
3
. The correlations between the raw scores are then
calculated as covariance between the standard scores N(a). Note
also that, just as in the EPR case, both the raw scores (1.33) and
the standard scores N(a) are manifestly non-contextual.
Equipped with these variables, we are now in a position to prove
the above theorem [4]. To this end, recall that no matter which model
of physics we are concerned withthe quantum mechanical model,
the hidden variable model, or any otherfor theoretical purposes all
we need to understand are the expectation values of the observables
measured in various states of the physical systems [10]. Accordingly,
consider an arbitrary quantum state [ 1, where 1 is a Hilbert
space of arbitrary dimensions, which may or may not be nite. We
impose no restrictions on either [ or 1, apart from their usual
quantum mechanical meanings. In particular, the state [ can be as
entangled as one may like, and the space 1can be as large or small as
one may like. Next consider a self-adjoint operator

O(a, b, c, d, . . . )
on this Hilbert space, parameterized by an arbitrary number of local
contexts a, b, c, d, etc. The quantum mechanical expectation value
of this observable in the state [ is then given by:
c
Q.M.
(a, b, c, d, . . . ) = [

O(a, b, c, d, . . . ) [ . (1.35)
More generally, if the system happens to be in a mixed state, then
c
Q.M.
(a, b, c, d, . . . ) = Tr
_
W

O(a, b, c, d, . . . )
_
, (1.36)
13
where W is a statistical operator of unit trace representing the state.
Our goal now is to show that this expectation value can always
be reproduced as a local-realistic expectation value of a set of binary
points of a parallelized 7-sphere. To this end, let
1 = A
a
() : IR
3
S
7
, 1 = B
b
() : IR
3
S
7
, etc.
(1.37)
be the raw scores of the form (1.33). Using prescriptions analogous
to (1.18) the corresponding standard scores then work out to be
N(a) : IR
3
S
6
, N(b) : IR
3
S
6
, etc. (1.38)
Here N(a), N(b), etc. may not necessarily be the same function for
all n IR
3
. They may be dierent functions for dierent directions.
This idealized prescription of raw scores and standard scores can,
and should, be further generalized. So far we have presumed that
randomness in these scores originates entirely from the initial state
representing the orientation of the 7-sphere. In other words, we
have presumed that the local interactions of the measuring devices
(J N(a) ) with the physical variables N(a) do not introduce
additional randomness in the scores A
a
(). Any realistic interaction
between (J N(a) ) and N(a), however, would introduce such a
randomness of purely local origin. We can model it by an additional
random variable = 1 with probability 0 p( [ a, ) 1 , so
that the bivectors (J N(a) ) representing the measuring devices
may now also take the random form ( J N(a) ). The average of
the corresponding raw scores A
a
() = 1 would then satisfy
1 A
a
() +1 , (1.39)
with A
a
() =

p( [ a, ) A
a
(, )
=
_

p( [ a, ) ( J N(a) )
_
(+ N(a) )
= (J N(a) ) (+ N(a) ) . (1.40)
Not surprisingly, this does not aect the corresponding standard
scores (1.38) worked out earlier. But if, in addition, we assume that
the common cause = 1 itself is also distributed non-uniformly
between its values +1 or 1, then the standard scores modify to
N(a) () N(a), N(b) () N(b) etc., (1.41)
14
where () =
1
_
1
_

_
2
, (1.42)
with being the average over the probability distribution of .
The correlation among the raw scores A
a
() = 1, B
b
() = 1,
C
c
() = 1, etc. can now be easily calculated as covariance among
the standard scores A
a
() = N(a), B
b
() = N(b), etc. as
c
L.R.
(a, b, c, d, . . . ) =
_

A
a
() B
b
() C
c
() . . . () d, (1.43)
where the overall probability distribution () =
m
() is allowed to
be both a non-uniform and continuous function of , with m being
the total number of local contexts a, b, c, d, . . . in the experiment.
To evaluate these correlations, note that the standard scores
A
a
() = N(a), B
b
() = N(b), C
c
() = N(c), etc. are in
fact bivectors representing the equatorial points of the 7-sphere,
which remains as closed under multiplication as the 3-sphere. As a
result, the product of the standard scores can be written as
A
a
() B
b
() C
c
() = f(a, b, c, . . . ) + P

N
() g (a, b, c, . . . ) ,
(1.44)
where the RHS is an octonionic spinor representing a non-equatorial
point of the 7-sphere, the vector

N(a, b, c, d, . . . ) IR
7
is a function
of all 3D vectors, P

N
()

N(a, b, c, d, . . . ) is a unit bivector
representing an equatorial point of S
7
that is dierent from the ones
represented by the bivectors A
a
(), B
b
(), C
c
(), etc., and the scalar
functions f(a, b, c, . . . ) and g(a, b, c, . . . ) satisfy f
2
+ g
2
= 1 with
f identied as the quantum mechanical expectation value (1.36):
f(a, b, c, d, . . . ) = Tr
_
W

O(a, b, c, d, . . . )
_
. (1.45)
Conversely, any arbitrary point of the 7-sphere (or joint beable)
( A
a
B
b
C
c
D
d
. . . )() = f(a, b, c, . . . ) + P

N
() g (a, b, c, . . . )
(1.46)
corresponding to the quantum mechanical operator

O(a, b, c, d, . . . )
can always be factorized into any number of local parts as
S
7
( A
a
B
b
C
c
D
d
. . . )() = A
a
() B
b
() C
c
() D
d
() . . . ,
(1.47)
since, as we have already noted, the 7-sphere remains closed under
multiplication of any number of its points. Using the identity (1.44),
15
the local realistic expectation value (1.43) can now be rewritten as
c
L.R.
(a, b, c, d, . . . ) = f(a, b, c, d, . . . )
_

() d
+ g (a, b, c, d, . . . )
_

N(n)
() () d. (1.48)
Note, however, that the vector

N(n) IR
7
appearing in the second
term here corresponds to a 3D direction n IR
3
that is necessarily
exclusive to all the other measurement directions a, b, c, d, . . . As
a result, the bivector P

N(n)
() =

N(n) necessarily corresponds
to a null measurement result, reducing the second integral in (1.48)
to zero [4][11]. If we next assume that the probability distribution,
although not necessarily uniform, remains normalized to unity,
_

() d = 1 , (1.49)
then the above expectation value reduces to
c
L.R.
(a, b, c, d, . . . ) = f(a, b, c, d, . . . ) . (1.50)
This nally proves our main theorem: Every quantum mechanical
correlation can be understood as a classical, local-realistic correlation
among a set of points of a parallelized 7-sphere. Q.E.D.
1.4 The Raison Detre of Quantum Correlations
The above result demonstrates that the discipline of parallelization in
the manifold of all possible measurement results is responsible for the
existence and strength of all quantum correlations. More precisely,
it identies quantum correlations as evidence that the physical space
we live in respects the symmetries and topologies of a parallelized
7-sphere. As we shall see in greater detail in Chapter 7, there are
profound mathematical and conceptual reasons why the topology
of the 7-sphere plays such a signicant role in the manifestation of
quantum correlations. Essentially it is because 7-sphere happens
to be homeomorphic to the most general possible division algebra.
And it is the property of division that turns out to be responsible
for maintaining strict local causality in the world we live in.
To understand this chain of reasoning better, recall that, just as a
parallelized 3-sphere is a 2-sphere worth of 1-spheres but with a twist
in the manifold S
3
(,= S
2
S
1
), a parallelized 7-sphere is a 4-sphere
worth of 3-spheres but with a twist in the manifold S
7
(,= S
4
S
3
).
16
More precisely, just as S
3
is a nontrivial ber bundle over S
2
with
Cliord parallels S
1
as its linked bers, S
7
is also a nontrivial ber
bundle, but over S
4
, and with entire 3-dimensional spheres S
3
as
its linked bers. Now it is the twist in the bundle S
3
that forces
one to forgo the commutativity of complex numbers (corresponding
to the circles S
1
) in favor of the non-commutativity of quaternions.
In other words, a 3-sphere is not parallelizable by the commuting
complex numbers but only by the non-commuting quaternions. In
a similar vein, the twist in the bundle S
7
,= S
4
S
3
forces one to
forgo the associativity of quaternions (corresponding to the bers
S
3
) in favor of the non-associativity of octonions. In other words,
a 7-sphere is not parallelizable by the associative quaternions but
only by the non-associative octonions. And the reason why it can be
parallelized at all is because its tangent bundle happens to be trivial:
TS
7
=
_
p S
7
p T
p
S
7
S
7
IR
7
. (1.51)
Once parallelized by a set of unit octonions, both the 7-sphere
and each of its 3-spherical bers remain closed under multiplication.
This, in turn, means that the factorizability or locality condition of
Bell is automatically satised within a parallelized 7-sphere. The lack
of associativity of octonions, however, entails that, unlike the unit
3-sphere (which is homeomorphic to the Lie group SU(2)), a 7-sphere
is not a group manifold, but forms only a quasi-group. As a result,
the torsion within the 7-sphere continuously varies from one point to
another of the manifold [4]. It is this variability of the parallelizing
torsion within S
7
that is ultimately responsible for the diversity and
non-linearity of the quantum correlations we observe in nature:
Parallelizing Torsion T

,= 0 Quantum Correlations.
The upper bound on all possible quantum correlations is thus set by
the maximum of possible torsion within the 7-sphere:
Maximum of Torsion T

,= 0 = The Upper Bound 2

2.
These last two results will be proved rigorously in Chapter 7.
1.5 Local Causality and the Division Algebras
In the last few sections we saw the crucial role played by the 3- and
7-dimensional spheres in understanding the existence of quantum
17
correlations. What is so special about 3 and 7 dimensions? Why
is the vector cross product denable only in 3 and 7 dimensions
and no other? Why are R, C, H, and O the only possible normed
division algebras? Why are only the 3- and 7-dimensional spheres
nontrivially parallelizable out of innitely many possible spheres?
Why is it possible to derive all quantum mechanical correlations as
local-realistic correlations among the points of only the 7-sphere?
The answers to all of these questions are intimately connected
to the notion of factorizability introduced by Bell within the context
of his theorem. Mathematicians have long been asking: When is a
product of two squares itself a square: x
2
y
2
= z
2
? If the number
z is factorizable, then it can be written as a product of two other
numbers, z = xy, and then the above equality is seen to hold for the
numbers x, y, and z. For ordinary numbers this is easy to check.
The number 8 can be factorized into a product of 2 and 4, and we
then have 64 = 8
2
= (2 4)
2
= 2
2
4
2
= 64. But what about sums
of squares? A more profound equality holds, in fact, for a sum of two
squares times a sum of two squares as a third sum of two squares:
(x
2
1
+ x
2
2
) (y
2
1
+ y
2
2
) = (x
1
y
1
x
2
y
2
)
2
+ (x
1
y
2
+ x
2
y
1
)
2
= z
2
1
+ z
2
2
.
(1.52)
There is also an identity like this one for the sums of four squares. It
was rst discovered by Euler, and then rediscovered and popularized
by Hamilton in the 19th century through his work on quaternions.
It is also known that Graves and Cayley independently discovered a
similar identity for the sums of eight squares. This naturally leads
to the question of whether the product of two sums of squares of
n dierent numbers can be a sum of n dierent squares? In other
words, does the following equality hold in general for any n?
(x
2
1
+x
2
2
+ +x
2
n
) (y
2
1
+y
2
2
+ +y
2
n
) = z
2
1
+z
2
2
+ +z
2
n
. (1.53)
It turns out that this equality holds only for n = 1, 2, 4, and 8. This
was proved by Hurwitz in 1898 [13]. It reveals a deep and surprising
fact about the world we live in. Much of what we see around us, from
elementary particles to distant galaxies, is an inevitable consequence
of this simple mathematical fact. The world is the way it is because
the above equality holds only for n = 1, 2, 4, and 8. For example,
the above identity is equivalent to the existence of a division algebra
of dimension n over the eld R of real numbers. Indeed, if we dene
vectors x = (x
1
, . . . , x
n
), y = (y
1
, . . . , y
n
), and z = x y in R
n
such
that z
i
s are functions of x
j
s and y
k
s determined by (1.53), then
[[x[[ [[y[[ = [[x y[[ . (1.54)
Thus the division algebras R (real), C (complex), H (quaternion),
and O (octonion) we use in much of our science are intimately related
18
to the dimensions n = 1, 2, 4, and 8. Moreover, from the equation
of a unit sphere,
x
2
1
+ x
2
2
+ + x
2
n
= 1 , (1.55)
it is easy to see that the four parallelizable spheres S
0
, S
1
, S
3
, and
S
7
correspond to n = 1, 2, 4, and 8, which are the dimensions of
the respective embedding spaces of these four spheres. What is not
so easy to see, however, is the fact that there is a deep connection
between Hurwitzs theorem and the quantum correlations [4]. As we
saw in the previous sections, all quantum correlations are inevitable
consequences of the parallelizability of the 7-sphere, which in turn is a
consequence of Hurwitzs theorem. So the innocent looking algebraic
equality (1.53) has far reaching consequences, not only for the entire
edice of mathematics, but also for that of quantum physics.
1.6 Concluding Remarks
The results obtained in the preceding sections (and in the chapters
that are to follow [4]) go far deeper and well beyond the boundaries
of Bells theorem. In any physical experiment what is observed are
clicks of the event detectors corresponding to yes/no answers to
our questions. As in the EPR experiment [9][12], when we compare
the answers recorded by various observers in quantum experiments
conducted at mutually remote locations, we nd that their answers
are correlated in a mathematically and statistically very disciplined
manner. The natural question then clearly is: why are these answers
correlated in such a disciplined manner when there appears to be no
predetermined common cause dictating the correlations. Bell and his
followers claimed that the observed correlations are the evidence of
a radical non-locality in nature. The stronger adherents of this view
often claim that nature is non-local. Here we have rejected this
view. Instead, we have demonstrated how a perfectly natural local
explanation of the observed correlations is possible. In our view these
correlations are the evidence, not of non-locality, but the fact that
the physical space we live in respects the symmetries and topologies
of a parallelized 7-sphere [4]. The 3-dimensional space we normally
presume as our reality is then simply one of the many bers of this
7-sphere. Our observations are still conned to various 3-dimensional
subspaces of this 7-sphere, but the correlations among the results of
our experiments are revealing that the observed measurement events
are actually occurring within a 7-dimensional manifold. The radical
non-locality of Bell is thus traded o for the extra dimensions going
beyond our immediate experiences in the macroscopic world.
The methodology that has led us to this conclusion is similar to
19
that used by Einstein to arrive at his local eld theory of gravity. Just
as the demand of locality in the face of Newtons non-local theory of
gravity led Einstein to general relativity, the demand of locality in
the face of quantum correlations has led us to a parallelized 7-sphere.
Acknowledgments
I wish to thank David Coutts, Fred Diether, Azhar Iqbal, Edwin
Eugene Klingman, Rick Lockyer, Ray B. Munroe, and Tom H. Ray
for discussions. I also wish to thank the Foundational Questions
Institute (FQXi) for supporting this work through a Mini-Grant.
Appendix 1: A 2-D Analogue of EPRB Correlation
Suppose Alice and Bob are two-dimensional creatures living in a
two-dimensional, one-sided world resembling a M obius strip, entirely
oblivious to the third dimension we take for granted (cf. Fig. 1.5).
Suppose further that they discover certain correlation between the
results of their observations that appears to be much stronger than
any previously observed correlation, and its strength appears to be
explainable only in non-local terms. Alice and Bob may, however,
strive to discover a hypothesis that could explain the correlation in
purely local terms. They may hypothesize, for example, that they are
in fact living in a M obius world, embedded in a higher-dimensional
space IR
3
. The purpose of this appendix is to illustrate how such a
hypothesis would explain their observed correlation in purely local
terms, and relate it to the hypothesis we have advanced above to
explain the correlations we observe in our three-dimensional world.
To this end, let Alice and Bob choose directions a and b to
perform two independent set of experiments, conducted at remote
locations from each other. Within their two-dimensional world the
vectors a and b could only have two coordinate components, but
Alice and Bob hypothesize that perhaps the vectors also have third
components, pointing outside of their own one-sided world. Let
the twisting angle between these external components be denoted by

ab
, as shown in Fig. 1.6 (b) below. Now the experiments Alice and
Bob have been performing are exceedingly simple. It so happens that
within their two-dimensional world wherever they set up their posts
and choose directions for their measurements, they start receiving
a stream of L-shaped patterns. Upon receiving each such pattern
they record whether it has a left-handed L-shape or a right-handed
L-shape. They determine this by aligning the longer arm of each
pattern along their chosen measurement direction, with the shorter
20
L
L
L
L
a
Figure 1.5: In the two-dimensional one-sided world of Alice and Bob
two congruent shapes may become incongruent relative to each other.
arm hanging in the opposite direction, as shown in Fig. 1.5. It is then
easy for them to see whether the pattern has a left-handed L-shape
or a right-handed L-shape. If it turns out to have a left-handed
L-shape, Alice and Bob record the number 1 in their logbooks,
and if it turns out to have a right-handed L-shape, they record the
number +1 in their logbooks. What they always nd in any such
experiment involving a large number of patterns is that the sum total
of all the numbers they end up recording independently of each other
always add up to zero. In other words, the L-shaped patterns they
both ceaselessly receive are always evenly distributed between the
left-handed patterns and the right-handed patterns. But when they
get together at the end of the day and compare the entries of their
logbooks, they nd that their observations are strongly correlated.
They nd, in fact, that the correlation among their observations can
be expressed in terms of the angle
ab
(dened above) as
c(a, b) = cos
ab
. (1.56)
To explain this correlation in purely local terms Alice and Bob
hypothesize that the three-dimensional space external to their own
is occupied by a mischievous gremlin, who is hurling complementary
L-shaped patterns towards them in a steady stream. What is more,
this gremlin has a habit of making a random but evenly distributed
choice between hurling a pattern towards his right or his left, with a
complementary pattern hurled in the opposite direction. Each choice
of the gremlin thus constitutes an evenly distributed random hidden
variable = 1, determining both the initial state of the patterns
21

ab
A
B
IR
3
(a)
z
z
a
b

ab
(b)
Figure 1.6: (a) The aerial view of the M obius world of Alice and Bob.
The distance between their observation posts is given by the angle
0
ab
2. (b) The cross-sectional view of the M obius world. The
torsional twist in the strip is characterized by the angle 0
ab
.
as well as the measurement outcomes of Alice and Bob:
A(a, ) =
_
+1 if = +1
1 if = 1
(1.57)
and B(b, ) =
_
1 if = +1
+1 if = 1 .
(1.58)
These outcomes, in addition to being local and deterministic, are also
statistically independent events. What is more, they are determined
manifestly non-contextually. Since the numbers A and B are read
o simply by aligning the received patterns along the chosen vectors
a and b and noting whether the patterns are right- or left-handed,
the values or directions of these vectors are of no real signicance for
the measurements of Alice and Bob. Any local direction a chosen
by Alice, for example, would give the same answer to the question
whether a received pattern is right-handed or left-handed. Moreover,
the z-components of a and b (inaccessible to Alice and Bob) also have
no signicance as far as their local measurements are concerned.
Now it may appear from the denitions (1.57) and (1.58) that
the product AB of the results of Alice and Bob will always remain
at the xed value of 1, but that, in fact, is an illusion. The product,
22
in fact, will uctuate inevitably between the values 1 and +1,
AB 1, +1 , (1.59)
not the least because A and B are statistically independent events.
To appreciate this, recall that in a one-sided world of a M obius strip
two congruent left-handed gures may not remain congruent for long
(cf. Fig. 1.5). If one of the two left-handed gures moves around the
strip relative to the stay-at-home gure it becomes right-handed,
and hence incongruent with the stay-at-home gure. And the same
is true for two right-handed gures
3
. This is quite easy to check by
making a model of a M obius strip from a strip of paper. If one starts
with two congruent L-shaped cardboard cutouts and moves one of
them around the strip relative to the other, then after a complete
revolution the two cutouts become incongruent with one another.
As a result the value of the corresponding product AB representing
congruence or incongruence of the two L-shaped gures would change
from +1 at the start of the trip to 1 at the end of the trip.
The reason for this of course is that there is a torsional twist in
the M obius strip similar to the one within the Cliord parallels that
constitute the 3-sphere [4]. This twist is what is responsible for the
strong correlation observed by Alice and Bob. To understand this
better, suppose the gremlin happens to hurl a right-handed pattern
towards Alice and a left-handed pattern towards Bob (i.e., suppose
= +1). What are the chances that Alice would then record the
number +1 in her logbook whereas Bob would record the number 1
in his logbook? The answer to this question would depend, in fact,
on where Alice and Bob are situated within the M obius strip, which
can be parameterized in terms of the external angle
ab
, as shown
in Fig. 1.6 (a). If the posts of Alice and Bob are almost next to each
other, then their patterns are unlikely to undergo relative handedness
transformation, and then Alice and Bob would indeed record +1
and 1, respectively, yielding the product AB = 1. If, however,
Bobs post is almost a full circle away from Alices post, then both
Alice and Bob would record +1 with near certainty, because then
Bobs pattern would have transformed into a right-handed pattern
relative to Alices pattern with near certainty, yielding the product
AB = +1. For all intermediate angles the probability of the two
patterns having undergone relative handedness transformation would
be equal to
ab
/2, and the probability of the same two patterns not
having undergone relative handedness transformation would be equal
to (2
ab
)/2. Thus all four possible combinations of outcomes,
++, +, +, and , would be observed by Alice and Bob, just
3
This is analogous to the real world fact that bivectors are not isolated objects
but represent relative rotations within the constraints of the physical space [4].
23
as in equation (1.26) discussed above. The corresponding correlation
among their measurement results would therefore be given by
c(a, b) = lim
n1
_
1
n
n

i =1
A(a,
i
) B(b,
i
)
_
=

ab
2
(AB = +1) +
(2
ab
)
2
(AB = 1)
=

ab
(+1) + (2
ab
) (1)
2
= 1 +
1

ab
_
i.e., linear within IR
3
_
. (1.60)
The validity of this result is straightforward to check by substituting

ab
= 0, , and 2 to obtain c(a, b) = 1, 0, and +1, respectively.
This correlation is expressed, however, in terms of the external
angle
ab
, which Alice and Bob can measure in radians as a distance
between their observation posts. As a nal step towards explaining
the correlation (1.56) they must now work out the distance
ab
in
terms of the angle
ab
, which is not a dicult task. Recalling the
properties of a M obius strip it is easy to see that these angles are in
fact related as

ab
= 1 cos
ab
. (1.61)
This relation is the dening relation of the M obius world of Alice
and Bob. Substituting it into equation (1.60) they therefore obtain
c(a, b) = 1 +
1

ab
= 1 +
1

[ 1 cos
ab
]
= cos
ab
. (1.62)
Needless to say, as enlightening as it is, this ctitious analogue of
the EPR-Bohm correlation cannot be taken too seriously. It helps us
to a certain extent in understanding the real world correlation within
the 3-sphere, but there are analogies as well as disanalogies between
the two worlds. For example, although a torsional twist within the
structure of both worlds is responsible for the sinusoidal correlation,
unlike the M obius strip a 3-sphere is an orientable manifold. Thus, it
is not the non-orientability, but the consistency of orientation within
the 3-sphere that brings about the variations ++, +, +, and
in the observed results of the real Alice and Bob. Moreover, unlike in
the M obius world where relative handedness of two L-shapes depends
on the distance between them, in the real world relative handedness
24
of the two bivectors appearing within the outcomes (1.11) and (1.12)
reects their intrinsic spinorial characteristics, independently of any
distance between them. It is constrained only by the identity
( a)( b) = a b (a b). (1.63)
This identity encapsulates the topology of a parallelized 3-sphere,
with = I specifying its orientation. It is clear from this identity
that when b = a the handedness (about a) of the bivectors on its
LHS dier from one another, whereas when b = +a the handedness
(about a) is the same regardless of the distance between a and b.
Appendix 2: The Meaning of a Geometric Product
The concept of a geometric product was rst introduced by Hermann
Grassmann to characterize what he called extensive magnitudes.
Nowadays Hestenes refers to extensive magnitudes as directed
numbers [5]. To understand the meaning of the geometric product
between two such directed numbers, let us rst look at the inner
product between two vectors, say a and b:
a b =
1
2
(ab + ba) = cos
ab
= b a, (1.64)
where
ab
is the angle between a and b. Clearly, this product is a
grade-lowering operation. It takes two grade-1 numbers, or vectors,
and gives back a grade-0 number, or a scalar.
Next, let us look at the outer product between a and b, as
dened by Grassmann:
a b =
1
2
(ab ba) = I (a b) = b a, (1.65)
where I (in the modern parlance) is a trivector. Unlike the previous
product this product is a grade-raising operation. It takes two grade-1
numbers, or vectors, and gives back a new directed number of grade
2; i.e., a bivector. Although an abstract entity, the bivector a b
may be visualized as an oriented plane segment, hovering orthogonal
to the vector a b.
Using the products (1.64) and (1.65), the geometric product
between the two directed numbers a and b can now be expressed as
ab = a b + a b. (1.66)
This product also takes two grade-1 numbers, or vectors, but gives
back an entity that is neither a scalar nor a bivector in general, but
25
rather a quaternion (or a spinor), made out of the grade-lowering
operation a b and the grade-raising operation a b. To appreciate
this, let us express the two components of the geometric product
(1.66) more explicitly as
ab = a b + I (a b)
= cos
ab
+ ( I c ) sin
ab
= exp( I c )
ab
, (1.67)
where c = a b/[a b[. It is now clear that the product ab is a
quaternion (or a spinor) that represents a rotation by an angle 2
ab
about the c-axis. The geometric product, ab, thus takes two grade-1
numbers, or vectors a and b, and gives back a pure act of rotation.
References
[1] A. Einstein, Dialectica 2, 320 (1948).
[2] A. Einstein, B. Podolsky, N. Rosen, Phys. Rev. 47, 777 (1935).
[3] J. S. Bell, Physics 1, 195 (1964).
[4] J. Christian, What Really Sets the Upper Bound on Quantum
Correlations?, arXiv:1101.1958
[5] D. Hestenes and G. Sobczyk, Cliord Algebra to Geometric Cal-
culus (Reidel, Dordrecht, 1984); T. G. Vold, Am. J. Phys, 61,
491 (1993); D. Hestenes, New Foundations for Classical Mechan-
ics, Second Edition (Kluwer, Dordrecht, 1999); D. Hestenes, Am.
J. Phys, 71, 104 (2003); C. Doran and A. Lasenby, Geometric
Algebra for Physicists (Cambridge University Press, 2003).
[6] A. Shimony, Brit. J. Phil. Sci. 35, 25 (1984).
[7] J. L. Rodgers and W. A. Nicewander, The American Statistician
42, 59 (1988).
[8] A. Aspect, P. Grangier, and G. Roger, Phys. Rev. Lett. 49, 91
(1982); See also A. Aspect, J. Dalibard, and G. Roger, Phys. Rev.
Lett. 49, 1804 (1982).
[9] G. Weihs et al., Phys. Rev. Lett. 81, 5039 (1998).
[10] J. von Neumann, Mathematical Foundations of Quantum Me-
chanics (Princeton University Press, Princeton, NJ, 1955).
[11] J. Christian, Disproofs of Bell, GHZ, and Hardy Type Theorems
and the Illusion of Entanglement, arXiv:0904.4259
[12] J. Christian, Disproof of Bells Theorem, arXiv:1103.1879.
[13] A. Hurwitz, Nachr. Ges. Wiss. Gottingen 309 (1898).
26